Video-to-script conversion method based on multi-stage AI cooperation and closed-loop self-healing

Abstract

The invention relates to a multi-stage AI cooperation and closed-loop self-healing video-to-script conversion method, and relates to the digital content creation cross technical field, the method comprises the following steps: obtaining a resource identifier ordered set of a series of videos, the resource identifier ordered set comprises a plurality of identifiers arranged in order, and each identifier corresponds to a video resource; based on the resource identifier ordered set, a preset first AI service model and a preset second AI service model, a first version script set is generated, the first AI service model represents a model with multi-mode video understanding and script generation capabilities, and the second AI service model represents a model with logic induction and information integration capabilities; creating a script verification object for each script in the first version script set, and performing closed-loop intelligent proofreading on each script verification object to generate a target script set. Through multi-stage AI cooperation and closed-loop self-healing, the technical effect of accurate, efficient and logically coherent video-to-script conversion is realized.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of artificial intelligence and digital content creation, and in particular to a video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing. Background Technology

[0002] With the booming development of the digital media industry, content production companies, broadcasting organizations, and streaming platforms have accumulated massive video asset libraries. These video asset libraries contain enormous value. Efficiently managing, archiving, retrieving, and reusing proprietary copyrighted content can unlock the secondary creation and derivative value of the content library, which is crucial for in-depth analysis of the content elements of a series. For example, because of the re-creation by directors, actors, etc., the final product of short dramas and long video series often differs from the script. When it is necessary to conduct in-depth analysis of an existing series of short dramas or long video series, or adapt it into a novel or audiobook, or produce multilingual subtitles, content summaries, and supplementary marketing materials, an accurate and structured script is required first.

[0003] Film and television companies digitize and structure scripts for classic dramas and conduct secondary creation; short video MCN agencies automatically generate plot scripts for operational data analysis; closed-loop self-healing during international subtitle generation ensures translation accuracy; and short video platforms automatically generate script summaries, etc.

[0004] This is crucial for a deeper understanding of the content elements of the series and for unlocking the secondary creation and derivative value of the content library.

[0005] Currently, some existing automated tools also attempt to solve the problem of converting videos to scripts. They usually adopt a simple linear processing mode, that is, to transcribe each video in isolation and generate corresponding script content for each video separately.

[0006] However, automated tools that employ simple linear processing models fail to capture the macro-narrative of a series when processed in isolation, leading to inconsistencies in character settings and key plot conflicts between different episodes, severely damaging the integrity of the story. AI models are prone to machine-generated flaws during the conversion process, such as garbled dialogue and repetitive descriptions, resulting in scripts that are not of quality suitable for professional production. Extensive manual post-production proofreading offsets the efficiency advantages of automation, making it unable to meet the needs of large-scale asset management. The one-way process cannot self-evaluate and correct errors, causing quality issues to be directly passed on to the downstream re-creation stage. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing, which aims to solve at least one of the above-mentioned technical problems.

[0008] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: Firstly, this application provides a multi-stage AI collaboration and closed-loop self-healing video-to-script conversion method, employing the following technical solution: A multi-stage AI collaboration and closed-loop self-healing video-to-script conversion method includes: Obtain an ordered set of resource identifiers for a series of videos, wherein the ordered set of resource identifiers includes multiple identifiers arranged in order, and each identifier corresponds to a video resource; Based on the ordered set of resource identifiers and the preset first AI service model, a set of initial script drafts is generated. The first AI service model represents a model with multimodal video understanding and script generation capabilities. Based on the set of initial script drafts and the preset second AI service model, a global logic attractor is generated, and the global logic attractor is used as a global constraint benchmark. The global logic attractor includes a story outline and a character list. The second AI service model represents a model with logical induction and information integration capabilities. Based on the ordered set of resource identifiers and the global logical attractor, a first version of the script set is generated; For each episode of the script in the first version script set, a script verification object is created, and a closed-loop intelligent proofreading is performed on each script verification object to generate a target script set. The closed-loop intelligent proofreading includes context-aware checking, guided regeneration, and arbitration.

[0009] The beneficial effects of this invention are as follows: Utilizing a first AI service model with multimodal video understanding and script generation capabilities, and a second AI service model with logical induction and information integration capabilities, a first version of the script set for the sequence videos is generated. Then, a target script set is generated through closed-loop intelligent proofreading, resolving the continuity issues in character and plot settings within the series of scripts, ensuring the integrity of the macro-narrative, constructing a closed-loop self-healing capability that automatically detects and repairs problems, guaranteeing the high quality and reliability of the final content, and enhancing the robustness and controllability of the system.

[0010] The script draft set is generated based on the ordered set of resource identifiers, the first AI service model, and unary constraints, ensuring that the core elements such as characters, plots, and scenes of the script drafts are consistent with those of the corresponding videos. A global logical attractor is generated based on the script draft set and the second AI service model and serves as the global constraint benchmark, solving the problem of continuity in character and plot settings in the series of scripts and ensuring the integrity of the macro narrative. The first version of the script set is generated based on the ordered set of resource identifiers, the global logical attractor, the first AI service model, binary constraints, and global constraints of the series of videos, ensuring the continuity of plots and the consistency of character states in adjacent script sets, further improving the overall quality and logic of the script set.

[0011] Based on the above technical solution, the present invention can be further improved as follows.

[0012] Furthermore, the process of generating a script draft set based on the ordered set of resource identifiers and the preset first AI service model includes: Create an independent script generation task for each identifier in the ordered set of resource identifiers; For each script generation task, based on the preset first AI service model and unary constraints, multimodal analysis is performed on the video resources corresponding to the identifier, and a first draft of the script is generated according to the preset script structure. The unary constraints represent the constraint that the first draft of the script is consistent with the core elements of the corresponding video. The core elements include characters, plot and scene. The set of first drafts of the script includes the first draft of the script corresponding to each video resource. For each script generation task, the quality compliance of the initial script draft is checked; For each script generation task, the first draft of the script that passes the quality compliance test will be stored in a set storage directory to generate a set of first draft scripts.

[0013] The beneficial effects of adopting the above-mentioned further solutions are as follows: creating an independent task for each identifier enables parallel processing and improves generation efficiency; combining the first AI service model and unary constraints to perform multimodal analysis of video resources and generate initial drafts according to a preset structure ensures that the initial drafts are consistent with the core elements of the corresponding videos; quality compliance testing of the initial drafts can screen out those that meet the requirements; storing qualified initial drafts in a set directory generates an initial draft collection, which facilitates subsequent processing and lays the foundation for generating the first version of the script set.

[0014] Furthermore, generating the first version of the play set based on the ordered set of resource identifiers and the global logical attractor includes: Based on the ordered set of resource identifiers of the series of videos, the global logical attractor, the preset first AI service model, binary constraints, and global constraints, a first version of the script set is generated. The binary constraints represent the constraints that ensure the coherence of plots and the uniformity of character states in adjacent script sets, and the global constraints represent the constraints that ensure the consistency of script content with the story outline and character list.

[0015] Furthermore, the step of creating a script verification object for each episode of the script in the first version of the script collection, and performing closed-loop intelligent proofreading on each script verification object to generate the target script collection includes: S31, Create a script verification object for each episode script in the first version script collection. The script verification object includes an identifier, a script, and a status. The initial status of each script verification object is pending inspection. S32, obtain the current state of each of the script verification objects in the current iteration, the current state including any one of pending inspection, inspection passed, pending repair, pending arbitration, repair failed, inspection in progress, repair in progress, and arbitration in progress; S33, among the script verification objects whose current state is to be checked in the current iteration, select one script verification object as the current script verification object; S34, Perform a script context-aware check on the current script verification object to obtain a structured diagnostic report of the current script verification object; S35, based on the structured diagnostic report of the current script verification object, perform guided rebirth on the current script verification object to obtain a new script and a new current state, update the current script verification object based on the new script, and execute S34 to S35 until the iteration number of the current script verification object reaches the set rebirth attempt number threshold or the current state of the current script verification object is checked and passed; S36, if the iteration count of the current script verification object reaches the set rebirth attempt threshold and the current state of the current script verification object is not checked and passed, then the current script verification object is arbitrated to obtain the final script, the current state of the current script verification object that has obtained the final script is updated to checked and the next iteration begins, executing S32 to S36 until all script verification objects whose current state is pending inspection are traversed. S37, Based on all states that the script corresponding to the script verification object has passed the check, generate the target script set.

[0016] The beneficial effects of adopting the above-mentioned further solution are as follows: Creating script verification objects for each episode of the script in the first version of the script collection and performing closed-loop intelligent proofreading achieves comprehensive quality checks and repairs of the scripts. Creating script verification objects facilitates the management of script identifiers, content, and status information; context-aware checks can detect errors in the scripts and generate structured diagnostic reports; guided rebirth can correct script errors based on the diagnostic reports; arbitration is conducted when the number of rebirth attempts reaches a threshold and still fails to pass the check, ensuring the generation of high-quality final scripts; finally, a target script collection is generated based on the scripts corresponding to all script verification objects that have passed the checks, ensuring the professionalism, accuracy, and usability of the final output scripts, building a closed-loop self-healing capability, solving the continuity problem of character and plot settings in the series of scripts, ensuring the integrity of the macro narrative, and enhancing the robustness and controllability of the system.

[0017] Furthermore, the step of performing a script context-aware check on the current script verification object to obtain a structured diagnostic report for the current script verification object includes: Determine whether the current state of the script verification object adjacent to the current script verification object is a stable state; If the current state of the script verification object adjacent to the current script verification object is a stable state, a structured diagnostic report is generated based on the preset third AI service model, the current script verification object, the scripts adjacent to the current script verification object, and the global logic attractor. The structured diagnosis includes the error state and the error cause. The third AI service model represents a model with content quality detection and error recognition capabilities. The determination of whether the current state of the script verification object adjacent to the current script verification object is a stable state includes: If the current state of the script verification object adjacent to the current script verification object is "Check passed", "Repair failed", or "Pending inspection", then the current state of the script verification object adjacent to the current script verification object is determined to be a stable state. Otherwise, determine that the current state of the script verification objects adjacent to the current script verification object is not a stable state.

[0018] The beneficial effects of adopting the above-mentioned further solution are as follows: When performing script context-aware checks on the current script verification object, it is first determined whether the state of adjacent script verification objects is stable, ensuring that the context information is reliable and does not change during the check, thus avoiding inaccurate check results due to the instability of adjacent script states; when the state of adjacent script verification objects is stable, a third AI service model with content quality detection and error identification capabilities is used to generate a structured diagnostic report by combining the current script verification object, adjacent scripts, and global logical attractors. This can accurately detect major machine errors in the script caused by the AI ​​transcription technology itself that are not of creative intent, clarify the error state and cause, and provide a basis for subsequent repairs.

[0019] Furthermore, the guided rebirth of the current script verification object based on the structured diagnostic report of the current script verification object to obtain a new script and a new current state includes: Based on the structured diagnostic report of the current script verification object, determine whether the current script verification object has a target error; If the current script verification object has a target error, the current status of the current script verification object is updated to "to be repaired". Based on the identifier of the current script verification object with the status of "to be repaired", the global logical attractor, the scripts adjacent to the current script verification object, the structured diagnostic report and the preset fourth AI service model, a new script corresponding to the current script verification object is generated. The fourth AI service model represents a model with guided content repair and generation capabilities. If the current script verification object does not have any target errors, then the current status of the current script verification object will be updated to "Check passed".

[0020] The beneficial effects of adopting the above-mentioned further solution are as follows: Based on the structured diagnostic report, it can accurately identify scripts that need repair by determining whether the current script verification object contains a target error. When a target error exists, the status is updated to "to be repaired," and a new script is generated using the fourth AI service model, combined with the script verification object identifier, global logical attractor, adjacent scripts, and the diagnostic report. This allows for targeted correction of errors in the script. If no target error exists, the status is updated to "check passed," avoiding unnecessary repair operations, ensuring the high quality and reliability of the final content, and building a closed-loop self-healing capability.

[0021] Furthermore, the arbitration of the current script verification object to obtain the final script includes: The status of the current script verification object is updated to pending arbitration. Based on the fifth AI service model, the identifier of the current script verification object, the global logical attractor, the scripts adjacent to the current script verification object, and all historical scripts corresponding to the current script verification object, the final script of the current script verification object is generated. The fifth AI service model represents a model with logical reasoning and global optimization decision-making capabilities.

[0022] The beneficial effects of adopting the above-mentioned further solution are as follows: When a script fails to be repaired after multiple rebirth attempts, its status is updated to pending arbitration. Using the fifth AI service model, which has logical reasoning and global optimization decision-making capabilities, and combining the identifier of the current script verification object, global logical attractors, adjacent scripts, and all historical scripts, a final script is generated. This realizes an automatic upgrade and resolution mechanism when conventional repair gets stuck in local optima, solves the problem of continuity in character and plot settings in a series of scripts, ensures the high quality and high reliability of the final content, enhances the robustness and controllability of the system, fundamentally solves the problem of continuity in character and plot settings in a series of scripts, and ensures the integrity of the macro narrative.

[0023] Secondly, this application provides a multi-stage AI collaboration and closed-loop self-healing video-to-script conversion system, which adopts the following technical solution: A multi-stage AI collaborative and closed-loop self-healing video-to-script conversion system includes: The acquisition module is used to acquire an ordered set of resource identifiers for a series of videos. The ordered set of resource identifiers includes multiple identifiers arranged in order, and each identifier corresponds to a video resource. The phased script generation module is used to generate a first version script set based on the ordered set of resource identifiers, a preset first AI service model and a preset second AI service model. The first AI service model represents a model with multimodal video understanding and script generation capabilities, and the second AI service model represents a model with logical induction and information integration capabilities. The closed-loop intelligent proofreading and self-healing module is used to create script verification objects for each episode of the script in the first version script set, and to perform closed-loop intelligent proofreading on each script verification object to generate the target script set.

[0024] Thirdly, this application provides an electronic device that adopts the following technical solution: An electronic device includes a memory and a processor, wherein the memory stores a computer program capable of being loaded by the processor and executing the video-to-script conversion method of multi-stage AI collaboration and closed-loop self-healing as described in any of the first aspects.

[0025] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the multi-stage AI collaboration and closed-loop self-healing video-to-script conversion method as described in any of the first aspects.

[0026] Additional aspects and advantages of this application will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of this application. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing, provided as an embodiment of the present invention. Figure 2 This is a schematic diagram of a closed-loop self-healing process for each script verification object provided in one embodiment of the present invention; Figure 3 A schematic diagram illustrating the overall solution process of a formal model provided in one embodiment of the present invention; Figure 4 A schematic diagram of the CSP solver is provided for one embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a video-to-script conversion system with multi-stage AI collaboration and closed-loop self-healing provided in one embodiment of the present invention. Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0030] This application provides a video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing. The method can be executed by an electronic device, which can be a server or a mobile terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The mobile terminal device can be a laptop, a desktop computer, etc., but is not limited to these.

[0031] The system of this invention is preferably deployed in a cloud server cluster or a distributed microservice architecture, supporting high concurrency processing and elastic scaling. Each functional module, including the acquisition module, the phased script generation module, and the closed-loop intelligent proofreading and self-healing module, can be deployed independently in a containerized or microservice manner, communicating through lightweight APIs to improve the maintainability and scalability of the system.

[0032] To more precisely define the technical solution of this invention in theory, we formalize the problem of converting a series of videos into a script as a constraint satisfaction problem (CSP). All subsequent implementation steps are built around this CSP model using a hierarchical iterative solver. The CSP model is defined as follows: Variables: Define a set of variables S = {S1, S2, ..., Sn}, where Si represents the final script content to be generated for the i-th episode of the video.

[0033] Domain: For each variable Si, its domain D(Si) is the set of all possible scripts that the AI ​​model can generate based on the video resource Ui.

[0034] Constraints: Cunary constraint: Cunary(Si,Ui) requires that the script content Si must be faithful to its corresponding original video content Ui.

[0035] Binary constraint (Cbinary): Cbinary(Si,Si+1) requires that the script content Si and Si+1 of adjacent episodes be logically consistent in terms of plot and character status.

[0036] Global Constraint (Cglobal): Cglobal(Si,G,C) requires that all script content Si must be consistent with the global "logical attractor" (i.e., the story outline G and the character list C).

[0037] The implementation system of this invention is essentially a hierarchical iterative solver of the aforementioned CSP, and its core operations are represented by the following functionalization: The formal constraints and convergence objective based on CSP are shown in the following formula: Logical attractor extraction: (G,C)=fattractor({fdraft(U1),...,fdraft(U1)} N )}).

[0038] Guided Rebirth:S i =fregen(U i ,S i-1 ,S i+1 ,(G,C),R i ), where R i For diagnostic reports.

[0039] Final Arbitration: S i (final) = farbiter(U i ,S i-1 ,S i+1 ,(G,C),H i ), where H i For all historical restoration evidence in episode i.

[0040] The global optimization objective of this solver is to find a set of solutions S* = {S_1*, ...,S_n*} that maximizes the number of scenarios that satisfy all constraints, i.e.: Maximize:Σ_{i=1 to N} I(C_unary(S_i)∧C_binary(S_i, S_{i+1})∧C_global(S_i))**; Here, I(...) is an indicator function, which has a value of 1 when all constraints within the parentheses are satisfied, and 0 otherwise. This optimization process is performed under the constraints of the number of rebirths (ci≤Cmax) and the total number of system loops (Tmain≤Tmax), ensuring the controllability and convergence of the solution process. The counters and loop control mechanisms in subsequent implementation processes all serve this purpose.

[0041] It should be noted that the global logic attractor consists of a story outline and serves as a global constraint benchmark for subsequent generation and proofreading, ensuring the macro-consistency of the script. The story outline includes the main plot and key turning points, while the character list includes character names, identity characteristics, and relationship networks.

[0042] To make the technical solution of this invention clearer, the core terms used in this specification are defined below: The global logic attractor representation refers to the global context benchmark extracted by the second AI service model through the analysis of all the initial script drafts in the series of videos. It includes the story outline (G) and the character list (C), which are used to constrain the subsequent script generation and repair process to ensure global logic consistency. A unary constraint represents the constraint that the core elements of the initial draft of the script must be consistent with those of the corresponding video. Binary constraints represent constraints that ensure the continuity of plot and the consistency of character states in adjacent episodes; Global constraints represent the constraints that ensure the script content aligns with the story outline and character list; In the closed-loop self-healing process, guided rebirth representation is based on diagnostic reports and global logical attractors, which call the fourth AI service model to specifically correct script errors and generate a new version. Arbitration represents the process of calling the fifth AI service model to make a final decision and generate a final script when conventional repairs fail. The entity state represents the state of an entity in the closed-loop intelligent verification and self-healing process. The entity state includes any one of the following: pending inspection, inspection passed, pending repair, pending arbitration, repair failed, inspection in progress, repair in progress, and arbitration in progress. The rebirth attempt threshold represents the maximum number of times an entity can perform closed-loop intelligent correction and repair. For example, the rebirth attempt threshold can be set to 3 to 5 times.

[0043] Based on the above CSP model, a preferred embodiment of this application has the following processing flow: like Figure 1 As shown, a multi-stage AI-based, closed-loop self-healing video-to-script conversion method includes: S1, Obtain an ordered set of resource identifiers for a series of videos, wherein the ordered set of resource identifiers includes multiple identifiers arranged in order, and each identifier corresponds to a video resource; In this embodiment, the acquisition device can be a network interface device, such as a network card, capable of reliably acquiring video resource identifiers from the network; or it can be a storage reading device, such as a hard disk reader, capable of reading identifiers from local storage. Upon startup, an ordered list containing multiple video resource identifiers is read from a specified storage location. This list defines the series of videos to be processed and their order.

[0044] S2, based on the ordered set of resource identifiers and the preset first AI service model, a set of initial script drafts is generated, wherein the first AI service model represents a model with multimodal video understanding and script generation capabilities; In this embodiment of the application, S2 specifically includes the following sub-steps: Create an independent script generation task for each identifier in the ordered set of resource identifiers; For each script generation task, based on the preset first AI service model and unary constraints, multimodal analysis is performed on the video resources corresponding to the identifier, and a first draft of the script is generated according to the preset script structure. For each script generation task, the quality compliance of the initial script draft is checked; For each script generation task, the first draft of the script that passes the quality compliance test will be stored in a set storage directory to generate a set of first draft scripts.

[0045] In the above embodiments, the electronic device receives an ordered set E = {U_1, U_2,..., U_N} of video resource identifiers, and creates an independent processing task for each identifier U_i. The task execution device can be a multi-core processor, which can process multiple tasks simultaneously to improve processing efficiency; or it can be a distributed computing node, where multiple nodes work together to speed up the generation process.

[0046] For each script generation task, the first AI service interface is invoked concurrently. This interface is configured with a general script generation instruction set, and the input includes video resource identifiers and preset script format requirements. Based on the preset first AI service model and unary constraints, multimodal analysis is performed on the video resource corresponding to the identifier, and a first draft script is generated according to the preset script structure. The first AI service model can be a deep learning-based multimodal model or a specially trained professional model focused on video understanding and script generation.

[0047] During multimodal analysis, the first AI service model extracts information such as audio content, visual features, and character actions from the video. Then, based on univariate constraints, it ensures that the initial script draft matches the core elements of the corresponding video. These core elements include characters, plot, and scenes. Next, the initial script draft undergoes quality compliance testing, which checks aspects such as length, content completeness, and logical coherence. Finally, the script drafts that pass the quality compliance test are stored in a designated storage directory, generating a collection of initial script drafts.

[0048] S3. Based on the set of initial script drafts and the preset second AI service model, a global logic attractor is generated, and the global logic attractor is used as a global constraint benchmark. The global logic attractor includes a story outline and a character list. The second AI service model represents a model with logical induction and information integration capabilities. In this embodiment of the application, S3 specifically includes: The electronic device integrates all the generated script drafts {D_1, D_2, ..., D_N} into a unified text data block in episode order; Call the second AI service interface, which has strong logical inductive ability. This interface is configured with a global information extraction instruction set (Prompt B), which is specifically used for macro analysis of cross-set content. Based on unified text data blocks and a pre-defined second AI service model, a global logical attractor for a series of videos is generated. The second AI service model can be a model with strong logical inductive ability, capable of analyzing and integrating a large amount of initial script content; or it can be an optimized information integration model focused on extracting global information. This model analyzes the overall narrative structure, identifies the main plot, key turning points, and timeline development, and extracts a coherent story outline; at the same time, it identifies and unifies the names, identities, characteristics, and relationship networks of all characters, constructing a standardized character list.

[0049] The story outline and character list are combined to form a global logical attractor, which serves as the global constraint benchmark for all subsequent generation and repair phases.

[0050] S4, Based on the ordered set of resource identifiers and the global logical attractor, generate a first version of the script set; In this embodiment of the application, generating a first version of the script set based on the ordered set of resource identifiers and the global logical attractor includes: Based on the ordered set of resource identifiers of the series of videos, the global logical attractor, the preset first AI service model, binary constraints, and global constraints, a first version of the script set is generated. The binary constraints represent the constraints that ensure the coherence of plots and the uniformity of character states in adjacent script sets, and the global constraints represent the constraints that ensure the consistency of script content with the story outline and character list.

[0051] Specifically, it includes: The electronic device reconstructs the processing request for each video resource identifier U_i, injecting a global logical attractor on top of the original input; Concurrent calls to the first AI service interface will cause the first AI service model to generate scripts with more coherent logic and more consistent character settings, while satisfying the constraints of video content fidelity and global logical consistency. The binary constraint requires that the plots of adjacent episodes be coherent and the character states be consistent, while the global constraint requires that the content of all scripts be consistent with the global logical attractor.

[0052] Guided by global constraints, a first-version script set is generated that is more logically coherent and has more unified character settings. The core of this method lies in transforming isolated single-episode generation into collaborative generation with global constraints.

[0053] S5, create a script verification object for each episode script in the first version script set, and perform closed-loop intelligent proofreading on each script verification object to generate a target script set. The closed-loop intelligent proofreading includes context-aware checking, guided regeneration, and arbitration.

[0054] In the embodiments of this application, such as Figure 2 As shown, the step of creating a script verification object for each episode of the script in the first version script collection, and performing closed-loop intelligent proofreading on each script verification object to generate the target script collection includes: S31, Create a script verification object for each episode script in the first version script collection. The script verification object includes an identifier, a script, and a status. The initial status of each script verification object is pending inspection. S32, obtain the current state of each of the script verification objects in the current iteration, the current state including any one of pending inspection, inspection passed, pending repair, pending arbitration, repair failed, inspection in progress, repair in progress, and arbitration in progress; S33, among the script verification objects whose current state is to be checked in the current iteration, select one script verification object as the current script verification object; S34, Perform a script context-aware check on the current script verification object to obtain a structured diagnostic report of the current script verification object; S35, based on the structured diagnostic report of the current script verification object, perform guided rebirth on the current script verification object to obtain a new script and a new current state, update the current script verification object based on the new script, and execute S34 to S35 until the iteration number of the current script verification object reaches the set rebirth attempt number threshold or the current state of the current script verification object is checked and passed; S36, if the iteration count of the current script verification object reaches the set rebirth attempt threshold and the current state of the current script verification object is not checked and passed, then the current script verification object is arbitrated to obtain the final script, the current state of the current script verification object that has obtained the final script is updated to checked and the next iteration begins, executing S32 to S36 until all script verification objects whose current state is pending inspection are traversed. S37, Based on all states that the script corresponding to the script verification object has passed the check, generate the target script set.

[0055] In this embodiment, the state of each script processing unit is maintained by the state management unit, the context-aware check unit ensures that the context is ready, the context-aware check unit calls the third AI service model to generate a diagnostic report, and the guided regeneration unit and the arbitration unit perform iterative repair under the management of the loop control unit until all scripts meet the quality requirements or reach the maximum number of iterations.

[0056] In this embodiment of the application, during guided regeneration, the regeneration operation is a specific implementation of the fregen function, which aims to generate a new solution Si′ with better constraint satisfaction for the constraint violation problem pointed out by the diagnostic report Ri; During arbitration, the arbitration process is a higher-order application of the farbiter function. When the regular regeneration gets stuck in a local optimum, the global optimum is sought by analyzing the complete case file Hi, which directly determines the final solution Si (final).

[0057] When describing the loop and counter, the rebirth limit Cmax and the system loop limit Tmax ensure that the aforementioned optimization objective converges within limited computational resources. In this embodiment of the application, optionally, the step of performing a script context-aware check on the current script verification object to obtain a structured diagnostic report of the current script verification object includes: Determine whether the current state of the script verification object adjacent to the current script verification object is a stable state; If the current state of the script verification object adjacent to the current script verification object is a stable state, a structured diagnostic report is generated based on the preset third AI service model, the current script verification object, the scripts adjacent to the current script verification object, and the global logic attractor. The structured diagnosis includes the error state and the error cause. The third AI service model represents a model with content quality detection and error recognition capabilities. The determination of whether the current state of the script verification object adjacent to the current script verification object is a stable state includes: If the current state of the script verification object adjacent to the current script verification object is "Check passed", "Repair failed", or "Pending inspection", then the current state of the script verification object adjacent to the current script verification object is determined to be a stable state. Otherwise, determine that the current state of the script verification objects adjacent to the current script verification object is not a stable state.

[0058] Optionally, the step of performing guided rebirth on the current script verification object based on the structured diagnostic report of the current script verification object to obtain a new script and a new current state includes: Based on the structured diagnostic report of the current script verification object, determine whether the current script verification object has a target error; If the current script verification object has a target error, the current status of the current script verification object is updated to "to be repaired". Based on the identifier of the current script verification object with the status of "to be repaired", the global logical attractor, the scripts adjacent to the current script verification object, the structured diagnostic report and the preset fourth AI service model, a new script corresponding to the current script verification object is generated. The fourth AI service model represents a model with guided content repair and generation capabilities. If the current script verification object does not have any target errors, then the current status of the current script verification object will be updated to "Check passed".

[0059] Optionally, the arbitration of the current script verification object to obtain the final script includes: The status of the current script verification object is updated to pending arbitration. Based on the fifth AI service model, the identifier of the current script verification object, the global logical attractor, the scripts adjacent to the current script verification object, and all historical scripts corresponding to the current script verification object, the final script of the current script verification object is generated. The fifth AI service model represents a model with logical reasoning and global optimization decision-making capabilities.

[0060] In the above implementation, the process of creating a script verification object for each episode of the script in the first version script set and performing closed-loop intelligent proofreading on each script verification object to generate the target script set includes the steps of creating a script verification object, iterative scheduling, checking, regeneration, arbitration, and generating the target script set.

[0061] The process involves creating script verification objects. The electronic device creates a script verification object for each episode in the first version of the script collection. These verification objects can be data structures that encapsulate information such as script content, current processing status, and repair history. Each script verification object is initially set to "pending inspection."

[0062] Iterative scheduling enters a main control loop limited by a maximum number of iterations. In each loop, the system traverses all script processing and verification objects and schedules tasks based on their status. The scheduling device can be a task scheduler, which can rationally allocate tasks based on the status of the script verification object; or it can be an intelligent algorithm engine, which performs task scheduling through optimization algorithms.

[0063] For script verification objects in the "Pending Inspection" state, a script context-aware check is performed. First, it determines whether the current state of script verification objects adjacent to the current script verification object is stable. Stable states include "Check Passed," "Repair Failed," and "Pending Inspection." The judgment device can be a state judgment module, judging the state of adjacent script verification objects according to preset rules; or it can be a logic judgment circuit, making judgments through hardware logic. If the states of adjacent script verification objects are stable, a structured diagnostic report is generated based on a preset third AI service model, the current script verification object, scripts adjacent to the current script verification object, and a global logic attractor. The third AI service model can be a model with content quality detection and error recognition capabilities, capable of multi-dimensional script detection; or it can be a specially trained detection model focused on discovering errors in the script. Detection dimensions include character consistency check, plot coherence check, context connection check, and machine-generated error check.

[0064] The methods for generating structured diagnostic reports include: The third AI service model categorizes and labels the discovered problems according to predefined error types and accurately locates the error locations; The third AI service model categorizes errors into three levels—minor, moderate, and severe—based on their impact on the overall script quality. Finally, a standardized diagnostic report is generated, which includes the error status, a detailed description of the problem, location information, and repair suggestions.

[0065] Rebirth: Based on the structured diagnostic report of the current script verification object, determine whether the current script verification object has a target error. If a target error exists, update the current status of the current script verification object to "Pending Repair," and generate a new script corresponding to the current script verification object based on the identifier of the current script verification object with the "Pending Repair" status, the global logical attractor, the adjacent scripts of the current script verification object, the structured diagnostic report, and the preset fourth AI service model. Call the fourth AI interface; the fourth AI service model can be a model with guided content repair and generation capabilities, which can specifically correct the script based on the error information; or it can be an optimized repair model focused on solving specific types of errors. If no target error exists, update the current status of the current script verification object to "Check Passed."

[0066] Targeted error correction methods include: Role consistency error correction: When a mismatch is detected in a role name or identity, the system will explicitly indicate the correct role setting in the repair instruction, guiding the AI ​​model to uniformly correct all related descriptions; Plot logic error correction: For plot points that conflict with the story outline, the system provides the correct development direction and constraints; Context break correction: For issues of inappropriate connection with adjacent set content, the system provides specific connection requirements and context information; Machine-generated error correction: For technical issues such as repetition and confusion, the system explicitly specifies the specific error patterns that need to be eliminated in the instructions.

[0067] Arbitration is performed if the current script verification object's iteration count reaches a set rebirth attempt threshold and its current state is not "passed." The Fifth AI service interface is invoked to update the current script verification object's state to "pending arbitration." Based on the Fifth AI service model, the current script verification object's identifier, global logical attractors, adjacent scripts, and all historical scripts corresponding to the current script verification object, a finalized script for the current script verification object is generated. The Fifth AI service model can be a model with logical reasoning and global optimization decision-making capabilities, capable of transcending local limitations and making decisions from a global perspective; or it can be a reinforced reasoning model focused on solving complex logical problems.

[0068] Generate the target script set based on the scripts corresponding to the scripts that have passed the script verification check in all states.

[0069] The formal model involved in this invention will be explained below.

[0070] To make the technical solution of the present invention clearer and easier to understand, this section will describe the core components of the aforementioned constraint satisfaction problem (CSP) model and its hierarchical iterative solver in conjunction with the accompanying drawings.

[0071] like Figure 3 and Figure 4 As shown, this invention formalizes the problem of converting a series of videos into a script into a CSP (Content Script Solving Method), and its overall solution process is reflected in... Figure 1 (Overall flowchart of the method) and Figure 2 In the (system module block diagram), Figure 1 It illustrates the complete high-level flow from input resource identifiers to the final generation of the target script set, covering the entire process of CSP solving. Figure 2 The system module block diagram shows the physical architecture for implementing the CSP solution, in which the "phased script generation module" and the "closed-loop intelligent proofreading and self-healing module" together constitute the hierarchical iterative solver.

[0072] Specifically, the function f_draft corresponds to Figure 3 The steps of "concurrent generation of the first draft of the script" and Figure 2 The "first draft generation unit" in the system, whose function is implemented by the first AI service interface, is responsible for generating an initial script based on a single video resource. The core logic of the function f_attractor is... Figure 1 The "Extracting the Overall Story Outline and Character List" section and Figure 4 The "Global Logic Attractor Extraction Unit" integrates all initial drafts through the second AI service interface and outputs logic attractors as global constraint benchmarks.

[0073] The specific execution of the function f_regen manifests as a "guided regeneration" step and a "guided regeneration unit," with the fourth AI service interface making targeted corrections to the diagnostic report. Meanwhile, the function f_arbiter, as the highest-level adjudication mechanism, is embodied in the "arbitration" stage and the "final arbitration unit," with the fifth AI service interface generating the final draft based on complete historical evidence.

[0074] Unary constraints, binary constraints, and global constraints are not implemented by a single module, but rather serve as core rules embedded in and permeating the call instructions of all the aforementioned core functions, guiding the behavior of all AI models.

[0075] The implementation principle of this embodiment is as follows: Through multi-stage AI collaboration, from initial draft generation to global logic refinement and final revision, the quality and logical consistency of the script are gradually improved. A closed-loop intelligent proofreading mechanism can promptly detect and correct errors in the script, preventing the accumulation of quality issues. Different AI service models are specifically optimized for different tasks, ensuring the processing effectiveness at each stage. This method improves the efficiency and quality of script generation, solves the problems of inconsistent global logic, low content quality, and lack of automatic repair mechanisms in existing technologies, and provides strong support for the management and reuse of content assets.

[0076] This method utilizes a first AI service model with multimodal video understanding and script generation capabilities, and a second AI service model with logical induction and information integration capabilities, to generate a first-version script set for sequential videos. Then, a closed-loop intelligent proofreading process is used to generate the target script set, resolving the continuity issues in character and plot settings within the series of scripts, ensuring the integrity of the macro-narrative, and constructing a closed-loop self-healing capability that automatically detects and repairs problems. This guarantees the high quality and reliability of the final content and enhances the robustness and controllability of the system.

[0077] Figure 5 A schematic diagram of the structure of a multi-stage AI collaboration and closed-loop self-healing video-to-script conversion system 200 is shown.

[0078] like Figure 5 As shown, a multi-stage AI collaboration and closed-loop self-healing video-to-script conversion system 200 includes: The acquisition module 201 is used to acquire an ordered set of resource identifiers for a series of videos. The ordered set of resource identifiers includes multiple identifiers arranged in order, and each identifier corresponds to a video resource. The first-stage script generation module 202 is used to generate a set of initial script drafts based on the ordered set of resource identifiers and the preset first AI service model. The first AI service model represents a model with multimodal video understanding and script generation capabilities. The second-stage script generation module 203 is used to generate a global logical attractor based on the set of initial script drafts and the preset second AI service model. The global logical attractor is used as a global constraint benchmark. The global logical attractor includes a story outline and a character list. The second AI service model represents a model with logical induction and information integration capabilities. The third-stage script generation module 204 is used to generate a first version script set based on the ordered set of resource identifiers and the global logical attractor. The closed-loop intelligent proofreading and self-healing module 205 is used to create script verification objects for each episode script in the first version script set, and perform closed-loop intelligent proofreading on each script verification object to generate a target script set. The closed-loop intelligent proofreading includes context-aware checking, guided rebirth, and arbitration. The final output and reporting module 206 is used to generate an audit report based on the target script set, and to organize each script in the target script set based on the audit report and script format to obtain the final script.

[0079] In one example, the module in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0080] For example, when modules in a device can be implemented via a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these modules can be integrated together as a system-on-a-chip (SOC).

[0081] In this application, various objects such as messages / information / devices / network elements / systems / apparatus / actions / operations / processes / concepts may be named. It is understood that these specific names do not constitute a limitation on the relevant objects. The names may be changed depending on the scenario, context, or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from their functions and technical effects embodied / performed in the technical solution.

[0082] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0083] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0084] Figure 6 This is a structural block diagram of an electronic device 300 according to an embodiment of this application.

[0085] like Figure 6 As shown, the electronic device 300 includes a processor 301 and a memory 302, and may further include one or more of an information input / output (I / O) interface 303, a communication component 304, and a communication bus 305.

[0086] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps in the aforementioned multi-stage AI collaboration and closed-loop self-healing video-to-script conversion method. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, as well as application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as one or more of Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0087] I / O interface 303 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 304 is used to test wired or wireless communication between electronic device 300 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 304 may include a Wi-Fi component, a Bluetooth component, and an NFC component.

[0088] The communication bus 305 may include a path for transmitting information between the aforementioned components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.

[0089] The electronic device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the video-to-script conversion method for multi-stage AI collaboration and closed-loop self-healing given in the above embodiments.

[0090] The following describes the computer-readable storage medium provided in the embodiments of this application. The computer-readable storage medium described below can be referred to in correspondence with the video-to-script conversion method of multi-stage AI collaboration and closed-loop self-healing described above.

[0091] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described multi-stage AI collaboration and closed-loop self-healing video-to-script conversion method.

[0092] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0093] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0094] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing, characterized in that, include: Obtain an ordered set of resource identifiers for a series of videos, wherein the ordered set of resource identifiers includes multiple identifiers arranged in order, and each identifier corresponds to a video resource; Based on the ordered set of resource identifiers and the preset first AI service model, a set of initial script drafts is generated. The first AI service model represents a model with multimodal video understanding and script generation capabilities. Based on the set of initial script drafts and the preset second AI service model, a global logic attractor is generated, and the global logic attractor is used as a global constraint benchmark. The global logic attractor includes a story outline and a character list. The second AI service model represents a model with logical induction and information integration capabilities. Based on the ordered set of resource identifiers and the global logical attractor, a first version of the script set is generated; For each episode of the script in the first version script set, a script verification object is created, and a closed-loop intelligent proofreading is performed on each script verification object to generate a target script set. The closed-loop intelligent proofreading includes context-aware checking, guided regeneration, and arbitration.

2. The video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing as described in claim 1, characterized in that, The process of generating a script draft set based on the ordered set of resource identifiers and the preset first AI service model includes: Create an independent script generation task for each identifier in the ordered set of resource identifiers; For each script generation task, based on the preset first AI service model and unary constraints, multimodal analysis is performed on the video resources corresponding to the identifier, and a first draft of the script is generated according to the preset script structure. The unary constraints represent the constraint that the first draft of the script is consistent with the core elements of the corresponding video. The core elements include characters, plot and scene. The set of first drafts of the script includes the first draft of the script corresponding to each video resource. For each script generation task, the quality compliance of the initial script draft is checked; For each script generation task, the first draft of the script that passes the quality compliance test will be stored in a set storage directory to generate a set of first draft scripts.

3. The video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing as described in claim 1, characterized in that, The process of generating a first version of the play set based on the ordered set of resource identifiers and the global logical attractor includes: Based on the ordered set of resource identifiers of the series of videos, the global logical attractor, the preset first AI service model, binary constraints, and global constraints, a first version of the script set is generated. The binary constraints represent the constraints that ensure the coherence of plots and the uniformity of character states in adjacent script sets, and the global constraints represent the constraints that ensure the consistency of script content with the story outline and character list.

4. The video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing as described in claim 1, characterized in that, The process of creating script verification objects for each episode of the script in the first version of the script collection, and performing closed-loop intelligent proofreading on each script verification object to generate the target script collection includes: S31, Create a script verification object for each episode script in the first version script collection. The script verification object includes an identifier, a script, and a status. The initial status of each script verification object is pending inspection. S32, obtain the current state of each of the script verification objects in the current iteration, the current state including any one of pending inspection, inspection passed, pending repair, pending arbitration, repair failed, inspection in progress, repair in progress, and arbitration in progress; S33, among the script verification objects whose current state is to be checked in the current iteration, select one script verification object as the current script verification object; S34, Perform a script context-aware check on the current script verification object to obtain a structured diagnostic report of the current script verification object; S35, based on the structured diagnostic report of the current script verification object, perform guided rebirth on the current script verification object to obtain a new script and a new current state, update the current script verification object based on the new script, and execute S34 to S35 until the iteration number of the current script verification object reaches the set rebirth attempt number threshold or the current state of the current script verification object is checked and passed; S36, if the iteration count of the current script verification object reaches the set rebirth attempt threshold and the current state of the current script verification object is not checked and passed, then the current script verification object is arbitrated to obtain the final script, the current state of the current script verification object that has obtained the final script is updated to checked and the next iteration begins, executing S32 to S36 until all script verification objects whose current state is pending inspection are traversed. S37, Based on all states that the script corresponding to the script verification object has passed the check, generate the target script set.

5. The video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing as described in claim 4, characterized in that, The step of performing a script context-aware check on the current script verification object to obtain a structured diagnostic report for the current script verification object includes: Determine whether the current state of the script verification object adjacent to the current script verification object is a stable state; If the current state of the script verification object adjacent to the current script verification object is a stable state, then a structured diagnostic report is generated based on the preset third AI service model, the current script verification object, the scripts adjacent to the current script verification object, and the global logic attractor. The structured diagnostic report includes the error state and the error cause. The third AI service model represents a model with content quality detection and error recognition capabilities. The determination of whether the current state of the script verification object adjacent to the current script verification object is a stable state includes: If the current state of the script verification object adjacent to the current script verification object is "Check passed", "Repair failed", or "Pending inspection", then the current state of the script verification object adjacent to the current script verification object is determined to be a stable state. Otherwise, determine that the current state of the script verification objects adjacent to the current script verification object is not a stable state.

6. The video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing as described in claim 4, characterized in that, The structured diagnostic report based on the current script verification object is used to perform a guided rebirth of the current script verification object to obtain a new script and a new current state, including: Based on the structured diagnostic report of the current script verification object, determine whether the current script verification object has a target error; If the current script verification object has a target error, the current status of the current script verification object is updated to "to be repaired". Based on the identifier of the current script verification object with the status of "to be repaired", the global logical attractor, the scripts adjacent to the current script verification object, the structured diagnostic report and the preset fourth AI service model, a new script corresponding to the current script verification object is generated. The fourth AI service model represents a model with guided content repair and generation capabilities. If the current script verification object does not have any target errors, then the current status of the current script verification object will be updated to "Check passed".

7. The video-to-script conversion method with multi-stage AI collaboration and closed-loop self-healing as described in claim 4, characterized in that, The arbitration of the current script verification object to obtain the final script includes: The status of the current script verification object is updated to pending arbitration. Based on the fifth AI service model, the identifier of the current script verification object, the global logical attractor, the scripts adjacent to the current script verification object, and all historical scripts corresponding to the current script verification object, the final script of the current script verification object is generated. The fifth AI service model represents a model with logical reasoning and global optimization decision-making capabilities.

8. A multi-stage AI collaboration and closed-loop self-healing video-to-script conversion system, characterized in that, include: The acquisition module is used to acquire an ordered set of resource identifiers for a series of videos. The ordered set of resource identifiers includes multiple identifiers arranged in order, and each identifier corresponds to a video resource. The first-stage script generation module is used to generate a set of initial script drafts based on the ordered set of resource identifiers and the preset first AI service model. The first AI service model represents a model with multimodal video understanding and script generation capabilities. The second-stage script generation module is used to generate a global logical attractor based on the set of initial script drafts and the preset second AI service model. The global logical attractor is used as a global constraint benchmark. The global logical attractor includes a story outline and a character list. The second AI service model represents a model with logical induction and information integration capabilities. The third-stage script generation module is used to generate a first version of the script set based on the ordered set of resource identifiers and the global logical attractor. The closed-loop intelligent proofreading and self-healing module is used to create script verification objects for each episode of the script in the first version script set, and to perform closed-loop intelligent proofreading on each script verification object to generate the target script set. The closed-loop intelligent proofreading includes context-aware checking, guided rebirth, and arbitration. The final output and reporting module is used to generate an audit report based on the target script set, and to organize each script in the target script set based on the audit report and script format to obtain the final script.

9. An electronic device, characterized in that, Includes a processor, which is coupled to a memory; The processor is configured to execute a computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Includes a computer program or instructions that, when run on a computer, cause the computer to perform the method as described in any one of claims 1-7.

Images